Unreal VFX (Niagara, Chaos destruction, VFX critique)

Expert level means an effect that tells the player the truth (where it hurts, when it is over), costs what its gameplay class allows across the whole world, and is proven by per-thread numbers on the target core count plus captures scored against a written rubric. Python cannot author Niagara emitters or modules (5.8 API): the agent works through templates with a User parameter contract, Material Instances, Effect Types and Dataflow recipes, and escalates the few graph edits as precise GUI requests. If a sibling skill named here is missing from your available skills, ask the user to install it (npx skills add scenario-labs/skills --skill <name>); unattended, proceed from tool schemas and flag the gap.

REQUIRED BACKGROUND: scenario-unreal-expert (channels, review loop, 5.8 traps).

Status (2026-09-24): Unreal is not installed. Offline checks pass (python3 tests/code/unreal-vfx/run_all.py --offline); every in-editor snippet is not yet run in Unreal; job_00_vfx_probe.py answers the [verify] names.

Stance (the expert delta)

• Gameplay truth first. The boldest edge sits on the collision boundary (missile head, AoE rim) and reads the gameplay number (User.Radius = damage radius), never a hand-matched constant; lingering says "still dangerous" (Keyser #1 [00:09:44], [00:15:21]).
• Contrast is a screen budget. Near-white core, never pure white; saturation in the fringe; dark saturated halo; contrast highest at the head; a weak projectile gets a hotter core and dark wisps, not a brighter trail (Keyser #3 [00:08:13], [00:12:03]; #4 [00:19:55]). Unexpected white is usually stacked additive duplicates: count spawns before touching color, then back the additive core with an alpha-blended saturated layer (Shishido [00:17:42]-[00:18:14]).
• Timing is a story. Anticipation rides the character's cast animation (no window: put the interest in the falloff); build-up stops spawning exactly at the hit; the flash peaks on the hit frame ("intensity is not size"); consequences follow causes (flash, then debris and sparks, then smoke); dissipation is shorter than build-up and ends inside the gameplay window. A fast round head strobes: stretch it along velocity and bake motion blur into its texture (Keyser #5 [00:08:13]-[00:09:54], [00:14:34], [00:18:07]; #2 [00:03:31]; Shishido [00:30:00], [00:37:02]).
• Optimize the world, not the effect. An Effect Type on every system first: Lyra went from 227 to 103 active systems, RT 5.07 to 1.51 ms, same look (Kiraly [00:20:48]). The Effect Type is set per system, never per emitter; secondary emitters get Scalability Mode Self. Read GT, GT Concurrent and RT separately at the target core count [00:05:15], then pick a lever that saves that thread: distance culling stops ticking, not rendering, and the scalability manager pays for the check; cull proxies save the tick, not the RT [00:16:57], [00:18:02]; every GPU system pays an RT dispatch even unseen [00:05:47]. FX budgets only on non-critical types, with fx.Budget.AdjustedUsageDecayRate against oscillation [00:36:50]-[00:37:21].
• Spawn architecture by class and rate (Kiraly's table [00:45:57]): pool moderate one-shots (Manual Release means Release to Pool), a Data Channel listener for spam, stateless emitters with system state off for simple bursts (footsteps GT 1,387 to 271 us [00:39:24]); NDC and stateless are exclusive; small counts stay on CPU.
• Destruction cost is active bodies and contacts. Fix in order: rendering (Nanite on the GC, root proxy with the ISM renderer), body count and lifetime (clusters, Remove on Break with random timers and Cluster Crumbling, a per-platform removal multiplier, Tiny Geo), collision (one-way debris, one convex per cluster, boxes under 0.3 relative size), throttles last (Caillaud wPgd1J1Tf70 [00:03:47]-[00:16:18]; Xiao Yue zFiHDRREv7E [00:29:27]).
• Weapons break walls with strain, not fields. Apply External Strain plus Apply Breaking Linear Velocity scaled by projectile speed, no strain falloff; fields are "heavy and slow" (Van Allen [00:29:29]). Pick one damage model: Material Strength and Connectivity (tune tensile strength on the physical material) or per-level thresholds (tune the array); size-specific is a flag, not a model (Xiao Yue [00:19:54]-[00:20:27]). Heft comes from friction (combine Max) and low restitution; stone density is 2 g/cm3, 2000 is a unit error (Van Allen [00:26:11]).
• Recipes, not clicks. Templates with exposed parameters and Dataflow graphs with per-asset variables survive art changes and are what an agent reaches reliably (Caillaud t_jyTILDYo8 [00:02:46]).

Establish first

Input	Why it changes the plan	Default when silent
Platform, frame rate, min-spec cores	per-thread budgets, quality tables	PC 60 fps; cores from scenario-unreal-performance, 4 if unknown (Kiraly)
Gameplay numbers (scenario-unreal-gameplay)	radius, speed, damage window, fire rate, max simultaneous, cast animation timing	ask; never guess a radius
Camera and environment	read at gameplay distance over the real ground colors	the game's camera and level
Style pillars, importance tier	brightness budget: idle, basic, defensive, damaging, game changer, ultimate (Keyser #1 [00:13:08])	three pillars from the art lead
Crowd and destruction scope	copies on screen; walls per level, must-break objects, low-end platforms	worst realistic fight; real time, Nanite on
Budgets	no source gives per-effect ms	[added] all Niagara at or under about 2 ms GT + GT Concurrent and 1.5 ms RT at min spec; confirm with scenario-unreal-performance

Workflow

1. Probe once per engine. job_00_vfx_probe.py (headless). GATE: probe.json answers the [verify] items.

2. Classify elements, group them into systems. classify_element per element, then system_plan: one Effect Type per system (its most critical element); the rest get Scalability Mode Self, or their own system when spammy or on GPU.

   System (examples)	Effect Type	Critical emitters	Secondary (Scalability Mode Self)
   Projectile, beam, tracer	NET_OneShotCritical	head	trail, wisps, embers
   Impact, hit, muzzle flash	NET_OneShotCritical	flash, damage ring	smoke, sparks (split if spammy or GPU)
   Debris or impact listener (NDC)	NET_OneShot, budget only with decay damping	none	all
   AoE, status, pickup	NET_PersistentCritical	rim	interior flavor
   Ambient dust, leaves, rain	NET_Ambient	none	all

   GATE: system_plan has no error.

3. Timing plan before building. Write the spawn schedule (wind-up from the cast animation, build-up ending at the hit, flash on the hit, debris and sparks, smoke, window end) and run spawn_schedule_verdict; strobe_risk(speed, head length) gives the stretch a fast head needs. GATE: no error; stretch in the template spec.

4. Template and contract. duplicate_template, user_parameters, contract_verdict; a missing control becomes one GUI request. Curves in that request pass fade_curve_verdict (opacity fades from mid-life, never shrink-only) and flipbook_playback_verdict (fast then slow). GATE: contract ok.

5. Look: materials and textures. Master material and instances from scenario-unreal-materials (erosion, gradient-map LUT, depth fade); reuse textures first (Trümpler [00:16:33]), generate glows, LUTs and noise in Python. Import masks with sRGB off, TEXTUREGROUP_Effects, big sources capped by Maximum Texture Size; baked flipbooks are premultiplied, so the material must divide RGB by A (premultiplied_check). Smoke and fire: baked flipbooks by default; Fluids are Beta. GATE: audit_vfx_textures, flipbook_check clean.

6. System properties and scalability. set_system_properties: effect_type, max_pool_size = the type's system cap, prime 0 for mid-play loads, warmup 0, current-frame data off for static effects; world-space trails need fixed bounds of speed x lifetime (fixed_bounds_check). Effect Types from EFFECT_TYPE_PRESETS (create_effect_type [verify] or GUI once). GATE: no NO_EFFECT_TYPE in audit_niagara; caps_verdict, budget_verdict clean.

7. Wire to gameplay. scenario-unreal-gameplay spawns with the pooling method, sets User parameters from gameplay data, spawns the wind-up from the cast animation, writes Data Channel payloads, applies strain on hit (strain_plan; collision plus strain for must-break objects). On hit, deactivate the trail and let its particles finish [added]. GATE: a PIE hit spawns from the pool (FX Outliner InUse / FreeInPool).

8. Destructible asset. Watertight, non-intersecting source (mesh_watertight_check); fixed seed; pattern by material (concrete noise and grout, wood shrink-fracture-stretch, glass slices or a crack material); two or three levels, anchored; Proximity graph with an area threshold; one damage model; one convex per cluster; interior material with a distance-to-surface bake (Caillaud t_jy [00:29:14]); Remove on Break ranges plus Cluster Crumbling; removal multiplier in the low-end device profile; disable settled debris that never wakes (Xiao Yue [00:28:45]). VFX on breaks needs Chaos data generation in Project Settings plus break notification on the component (enable_break_events, Xiao Yue [00:08:42]); switch only broken bones to a debris profile. Then harden_gc; piece count in a gym at 10, 50, 100 (Caillaud t_jy [00:35:12]). GATE: gc_recipe_verdict, gc_verdict, physmat_check clean.

9. Review loop. Textured review level, Manual exposure, the gameplay camera; live_capture_series.py or capture_ages with clean plates; score_captures.py (a blank frame voids the series); LOOK at the color, grayscale and squint sheets; repeat in PIE ("the viewport lies", Shishido [00:05:05]), over green and brown ground for the color-blind check, in a crowd grid and at sg.EffectsQuality 0 to 3. On white-out run duplicate_spawn_check on Debugger counts first. One change per pass. GATE: no rubric fail; judgment rows scored (references/critique.md).

10. Measure. Fixed worst fight at the target core count: stat unit, stat NiagaraOverview, Debug HUD, Insights with -statnamedevents (Niagara is invisible without it, Kiraly [00:10:52]); init_for_performance_baseline(); lever_check before each change; budget_oscillation when budgets are on. Destruction: frame series around the break, CVD, t.OverrideFPS 60 [verify]. GATE: niagara_overview_verdict, before_after, destruction_run_verdict within budget, visual parity.

11. Deliver the budget sheet to scenario-unreal-performance.

Numbers

Value	Relative to	Source
NET_OneShot: 15 per system, about 30 per type, Spawn Only, Kill and Clear, instant cull 0.0	fire-and-forget, not critical	Kiraly [00:22:26]
NET_PickupPersistent: 5 (sight lines), Continuous, Asleep, cull delay 1.0 s	persistent critical	Kiraly [00:24:03]
Max Pool Size 32, Prime 0 (defaults); type cap = pool size	per system	Kiraly [00:32:18], [00:33:22]
Flipbook 8 x 8 on 1024 = 128 px tiles, end frame 63, bake at 30 fps	baker	Flipbook Baker doc
Thresholds 5000 / 500 / 50; strain 50000, radius 100, velocity 500	quickstart wall	Chaos docs
Boxes below 0.3 relative size; 4,000 fragments in 4 levels, about 200 bodies hit	leaf proxies; clustering	Xiao Yue [00:25:51], [00:22:32]
Remove on sleep 3 s; removal 5 to 10 s, then a per-platform multiplier	debris	Caillaud [00:10:43], [00:11:24]

Quality gates

• Measurable: audits (audit_niagara, audit_gc, audit_vfx_textures, caps_verdict, system_plan, gc_recipe_verdict); per-thread stats at the target core count, averages and spikes; break spike and recovery; capture rubric (footprint vs radius, value bands, blowout, saturation under values, 60 / 30 / 10 with the accent at the focal point, size classes held over time, head contrast, color-blind contrast, intensity curve, schedule, strobing); image_checks rejects blank frames.
• Visual: grayscale read at gameplay distance; popcorning; story beats; crowd read; no pops across quality levels; the break reads as the material with thick interiors, no floating chunks, heavy debris, removal hidden by dust.

Common mistakes

Mistake	What it looks like	Fix
Hand-matched radius	ring larger than the damage zone	User.Radius from gameplay
White blowout	flat white blob	count duplicates, then an alpha-blended saturated layer behind the additive core
Round head at speed	strobing dots	stretch along velocity, motion blur in the texture
Build-up after the hit; smoke before the flash	muddy payoff	spawn schedule: stop at the hit, causes first
Shrink-only fade	popcorning cards	opacity fade from mid-life, matched layer colors
Effect Type per emitter	embers get the uncullable type	one type per system, Scalability Mode Self
Distance culling to cut RT	RT unchanged	lever_check: caps, CPU sims, stateless
Budget without decay	cull, respawn, cull	decay rate, non-critical types only
Fixed bounds smaller than the trail	trail vanishes on screen	fixed_bounds_check
Destroying the projectile on impact	trail pops off	deactivate, let particles finish
Sleep to hold a wall	wakes on a bump	anchor or kinematic bones
Default connection graph	floating chunks, corner hinges	Proximity or Bounds Overlap Filtered
MSC with a tuned threshold array	tuning does nothing	tensile strength on the physical material
Field or strain falloff per bullet	physics spike	strain plus breaking velocity, no falloff
Debris profile on the whole GC	player walks through the intact wall	switch broken bones on the break event
Sleeping debris that never wakes	solver cost	disable
gc.removal_duration = ...	refused (read-only attribute)	set_editor_property

Handoffs

• Receives from scenario-unreal-materials: particle master material, instances, texture settings; from scenario-unreal-gameplay: numbers, cast animation timing, pooled spawns, NDC writes, strain calls; from scenario-unreal-world-building: destructible placements and counts.
• Delivers to scenario-unreal-performance: a budget sheet (system, Effect Type, caps, pool, sim target, per-thread cost, core count, quality; GCs with pieces, peak active bodies, break spike, recovery) plus the repeatable scenario; to scenario-unreal-cinematics: Chaos caches, Cinematic scalability; to scenario-unreal-lighting-rendering: particle lights, Heterogeneous Volumes. Cloth belongs to scenario-unreal-animation.

UE 5.8 notes

• Data Channels Production Ready since 5.5 (the GDC 2025 talk still says experimental); lightweight emitters Beta since 5.5, compile step dropped in 5.7; Niagara Fluids Beta; "Empty" template is "Minimal".
• r.HeterogeneousVolumes.Allow (5.8); Niagara Editor Preview Actor (5.7).
• Chaos: default Dataflow template at GC creation, runtime Dataflow evaluation (5.8); destruction throttles in Project Settings (5.6).
• GC Python: root_proxy_data, size_specific_data, set_dataflow_asset; per-bone data (anchors, Remove on Break, one-way levels) is not on the class.

References

• references/expert-notes.md: principles by expert, timestamps, disagreements, what the U7 grade found missing.
• references/procedures.md: full code, test path and status per procedure.
• references/critique.md: the rubric (metrics, thresholds, judgment rows).
• references/gui-paths.md: editor menus for the same work.
• references/sources.md: every source, credential, URL, best timestamps, revision history.
• scripts/ue_vfx.py: offline verdicts and metrics; in-editor helpers (not yet run).